Both The Scientist and Nature feature scientific misconduct prominently in their current issues:
A Fluctuating Reality (The Scientist)
Breeding Cheats (Nature - see also related articles)
An earlier Wired article also covers the issue:
Faked Research Results on Rise?
My take on this is that the incidences of scientific fraud will continue to increase in the coming decade(s). While each case is founded in its own particular circumstances, the general situation of raised stakes in science will contribute statistically to an overall increase in scientific fraud.
This ties in nicely with the debate on the number of scientists being trained:

• Are we training too many scientists? (The Scientist)
• Are There Too Many Postdocs? (Science)
• Too Many or Too Few? The Postdoc Production Policy Debate (Science)

With increased numbers of scientists and decreased funding, competition rises. Currently, every single scientist on this planet feels the pressure that he/she needs to become a science superstar in order to survive and obtain a position which will pay the bills. A superstar will only be born in a fashionable topic and thus these topics (largely controlled by a few science journals) are overrun, increasing competition further. Obviously, only very few will become science superstars. Consequently, the incentives of behaving fraudulently have never been larger than today. The number of fraud cases will inevitably follow this trend. Because of the huge incentives (getting a job and fame or landing on the streets in disgrace) I'm doubtful that any control measures can stop these parallel developments.

However, reducing the incentives on the high end (less fame and prestige, less spin-off companies, patents and luxurious meetings sponsored by drug companies) and cushioning the low end (e.g. by capping grant size to increase overall grant number) will also decrease the number of fraud cases in science.

If you are a PostDoc with a family, your contract runs out in three months and every faculty position has 300 applicants, you really feel the temptation to fiddle a little with this one graph which will get you the publication you need to beat the other 299 in order to feed your family. Asking for honesty is probably rather ineffective in such a situation.
That's the much more common low-profile fraud which is probably not increasing but exploding even as I'm typing this.
The reasons are clear. Are we going to do something about it?

Confirming my thoughts, the Nature article cited above supports the "pressure cooker" notion of breeding cheats:

So obviously, the way I was treated by the DFG provides a great example of how funding agencies should not behave...